Jin Huang (email@example.com) is an Assistant Professor at School of Social Work, Saint Louis University and Yunju Nam (firstname.lastname@example.org) is an Assistant Professor at School of Social Work, University at Buffalo, State University of New York. Margaret Sherrard Sherraden (email@example.com) is Professor at School of Social Work, University of Missouri-St. Louis. Support for the SEED for Oklahoma Kids experiment comes from the Ford Foundation, Charles Stewart Mott Foundation and Lumina Foundation for Education. We value our partnership with the State of Oklahoma, including State Treasurer Ken Miller, former State Treasurer Scott Meacham, Tim Allen, James Wilbanks, Kelly Baker, Derek Pate, Sue Mallonee, Tony Mastin and James Conway. We appreciate the contributions of Ellen Marks, Bryan Rhodes and Jun Liu at RTI International. The authors thank Margaret Clancy for her careful review and insightful comments on the manuscript. We are grateful to Michael Sherraden, Sandy Beverly and Mark Schreiner for their valuable discussions on the SEED OK project. The authors thank Youngmi Kim and Nora Wikoff for their support on data management. Erin Moriarity provided valuable research assistance. Carrie Freeman, Tiffany Trautwein and Julia Stevens provided editing assistance.
This study examines how study participants' financial knowledge and participation in a Child Development Account intervention affects 529 College Savings Plan account holding among caregivers of infants. The study uses data from the SEED for Oklahoma Kids (N = 2,651), a statewide randomized experiment using a probability sample of infants selected from birth records. Results of logit regression show that participants' financial knowledge is positively related to account holding in the treatment group but not in the control group. The interactive effects between financial knowledge and treatment status are statistically significant. This finding implies that the effect of financial knowledge on financial decisions related to college savings is moderated by institutional features, such as incentives, information and access. Results of this study support the propositions of financial capability and suggest that expanding financial capability requires both improved individual financial knowledge and supportive policy.
Awareness of the importance of financial capability in optimal financial decision making and financial well-being is growing (Johnson and Sherraden 2007; Sherraden in press). To be financially capable, families must have not only financial knowledge and skills but also access to appropriate financial products and services. Among low-income and low-education families, financial capability is especially important because financial literacy tends to be lower (Mandell 2008) in a context of more acute financial challenges and lower access to financial services (Bucks et al. 2009).
The concept of financial capability has three key propositions: (1) financial knowledge and skills are important determinants of an individual's financial ability and well-being, (2) individuals require access to financial institutions, and (3) there are interactive effects between the two. An individual's financial ability is influenced by the institutional setting that creates (or constrains) certain opportunities to apply financial knowledge and skills. The institutional setting is part of what Martha Nussbaum refers to as a person's “external conditions” (2000, 85). Together, individual ability and institutional setting shape a person's financial capability and well-being.
This study tests the three propositions regarding financial capability using data from the SEED for Oklahoma Kids (SEED OK). SEED OK is a statewide randomized social experiment to test a Child Development Account (CDA) program that encourages families to accumulate assets for their children's future using the existing 529 College Savings Plan in Oklahoma (Oklahoma College Savings Plan or OK 529 plan). CDAs are savings accounts for children that provide a structured opportunity to save and accumulate assets by providing incentives, information and access (Cramer and Newville 2009; Sherraden 1991). 529 College Savings Plans are tax-advantaged savings programs designed by the federal government and operated by state governments to encourage saving specifically for future college costs (Clancy 2003; U.S. Department of Treasury 2009).
The SEED OK experiment offers various college savings options to participants, which provides a unique opportunity to examine the interaction between individual ability and policy environments. Specifically, this study uses 529 account-holding status as an outcome measure, and investigates three research questions: (1) Does financial knowledge increase an individual's chance of holding a 529 account for a child's future? (2) Does a CDA intervention (i.e., the SEED OK experiment) increase an individual's likelihood of 529 account holding regardless of level of financial knowledge? (3) Are there any interactive effects between an individual's financial knowledge and the CDA intervention? That is, does financial knowledge have a different impact on the probability of holding a 529 account for the SEED OK participants in the treatment and control groups?
Financial Literacy and Financial Capability
Research on financial literacy has grown substantially in recent years (Lusardi and Mitchell 2011; U.S. Financial Literacy and Education Commission 2007). Financial literacy has been proposed as an effective approach for individuals to achieve optimal financial decisions. Financial literacy is defined as an individual's knowledge and skills needed for efficient management of that individual's financial resources (Huston 2010; Remund 2010). Research shows that financial literacy is a key determinant of an individual's financial functioning (e.g., spending, saving, borrowing and investing) (Behrman et al. 2010). The effects of financial literacy on financial functioning are retained even after controlling for an individual's socioeconomic characteristics (Lusardi and Mitchell 2011), and the importance of financial literacy on financial decision making has been confirmed even in experimental settings (Agnew and Szykman 2011). Therefore, appropriate financial education and training should be provided for individuals to improve their financial knowledge, especially for populations with low socioeconomic status and education.
Less is known about financial capability than financial literacy, despite increased research and policy interest in this topic. While the concept of financial literacy focuses on individual ability (e.g., financial knowledge and skills), the concept of financial capability suggests that access to appropriate financial services (i.e., the institutional setting) is also critical for building financial well-being. In this way, financial capability considers both an individual's ability to act (based on knowledge and skills) and opportunity to act (through access to appropriate and beneficial financial services) (Johnson and Sherraden 2007; Sherraden in press).
In practice, the concept of financial capability has been applied to multiple innovations in financial products and services (Sherraden in press). From the perspective of financial capability, promoting savings—such as college savings for children—requires improving not only financial knowledge and skills through education and training but also access to financial services and the ways these services are designed and delivered. Regarding promoting access to financial services, Sherraden and colleagues have proposed an institutional theory of saving and identified seven constructs that shape household saving behavior: incentives, information, access, facilitation, expectations, restrictions and security (Sherraden 1991; Sherraden and Barr 2005; Sherraden, Schreiner, and Beverly 2003)
In summary, financial capability does not reside solely within the individual; it captures the relationship between people's internal ability and their external conditions and constitutes what Amartya Sen suggests is access to “real opportunities” (1999, 75). The following two sections discuss financial capability from the aspects of internal ability and external conditions using the SEED OK experiment as an external condition within which study participants make financial decisions on college savings.
Financial Knowledge as a Measure of Individual Ability
Financial capability is difficult to measure because it incorporates individual and institutional features. Financial knowledge—which refers to an individual's understanding of financial concepts—is an indicator of individual ability. It has been commonly used as a measure or a proxy for financial literacy. In fact, financial literacy and financial knowledge are often used interchangeably. Huston (2010) summarizes measures of financial knowledge in 71 individual studies drawn from 52 different data sets and identifies four distinct content areas in the definition of financial knowledge: (1) basic money concepts, (2) borrowing, (3) saving or investment and (4) protection concepts. In most cases, financial knowledge is measured by multiple items, and a threshold value or a grading system is provided to explain levels of financial knowledge (Huston 2010).
Previous research links financial knowledge to financial functioning (e.g., saving and portfolio choice). For example, lack of financial knowledge or low financial literacy is found to be negatively related to retirement planning, saving (Behrman et al. 2010; Lusardi and Mitchell 2006, 2008, 2011), wealth accumulation and stock investment (Christelis, Jappelli, and Padula 2010; van Rooij, Lusardi, and Alessie 2007). Low financial knowledge also is associated with paying higher interest rates and fees (Lusardi and Tufano 2009) and higher rates of housing delinquency, default and foreclosure (Gerardi, Goette, and Meier 2010). Individuals with low levels of financial knowledge are less likely to select mutual funds with lower fees (Hastings and Tejeda-Ashton 2008) and are less likely to diversify their investment portfolios (Guiso and Jappelli 2008).
Few studies focus on the impact of financial knowledge on financial planning for children's college. A recent survey (Sallie Mae 2010) shows that saving for college has become as high a priority as saving for retirement for U.S. families with children. Even so, about 70% of low-income families have not saved for their children's college education. This study, with a focus on financial capability, addresses this knowledge gap by looking at parents' financial knowledge related to college savings in a Children's Development Account (CDA) intervention.
The SEED OK Experiment: An Institutional Intervention
SEED OK is a statewide randomized policy experiment of CDAs to encourage families to accumulate savings for their children's future education. The overall purpose of the SEED OK experiment is to test a universal and progressive policy of lifelong asset building beginning at birth. The SEED OK experiment changes the institutional setting that study participants face when making financial decisions related to college savings for their child. SEED OK tests whether and how a CDA intervention encourages families to open a college savings account and save for a child's college education and promotes social and economic development.
Existing studies show household assets—especially savings for college education—have a positive association with children's educational attainment and other development outcomes (Conley 2001; Elliott and Beverly 2011; Lerman and McKernan 2008; Nam, Huang, and Sherraden in press). Likely this is a result of helping families pay for postsecondary education, motivating young people to prepare for college, and offering young people opportunities for learning about financial management (Nam, Huang, and Sherraden in press; Sherraden 1991).
SEED OK is built on the existing account structure of the OK 529 plan, which provides tax incentivized accounts for college savings (Zager et al. 2010). The experiment is a partnership of the State of Oklahoma (Treasurer's Office, Department of Health, Department of Human Services, Tax Commission and Oklahoma College Savings Plan), the Center for Social Development at Washington University in St. Louis, and RTI International (Nam et al. in press; Zager et al. 2010).
SEED OK randomly assigned 2,704 infants into the treatment and control groups and offered additional financial incentives for and information about the OK 529 plan to treatment participants in addition to existing tax benefits (see Figure 1). First, all treatment participants received a $1,000 initial deposit in a state-owned 529 account, which was opened automatically for each treatment child unless the participant opted out of this option. Parents were not given an option to open a state-owned account on their own and were not allowed to make deposits into this account. The OK 529 plan does not include state-owned accounts, so this is a unique feature of SEED OK.
Second, treatment participants were encouraged to open their own participant-owned 529 accounts, the typical savings account offered by state 529 plans. Participant-owned accounts are important savings vehicles because participants' contributions are allowed. Treatment group participants had the opportunity to receive a time-limited $100 incentive for opening participant-owned accounts by April 15, 2009.
Third, income-eligible participants in the treatment group were offered a savings match for deposits made in participant-owned accounts. Households with an annual adjusted gross income below $29,000 were offered a 1:1 match (i.e., for every dollar a participant deposited, SEED OK would match one dollar), and households with an annual adjusted gross income from $29,001 to $43,499 were eligible for a 0.5:1 match (i.e., for every dollar a participant deposited, SEED OK would match 50 cents).
Members of the control group did not receive any information from SEED OK about the OK 529 plan, were not eligible for the state-owned account, and were not offered any SEED OK financial incentives. However, they could open their participant-owned accounts in the OK 529 plan, just as any non-study participant could (see Figure 1). In summary, the SEED OK experiment expanded the existing OK 529 plan to include three institutional features: incentives (e.g., $100 account-opening incentive and deposit matches), information and access (through automatic account opening).
The SEED OK experiment created options for participants to save for their children's college education: (1) treatment and control group participants had the choice to open OK 529 plan participant-owned accounts and (2) treatment participants could receive an additional account-opening incentive and savings matches for their participant-owned accounts. Therefore, participants' financial decisions for their participant-owned accounts provide a unique opportunity to examine the interaction between financial knowledge and policy features.
In particular, we have three hypotheses corresponding to the three research questions stated in the Introduction: (1) Hypothesis 1: participants' financial knowledge increases the probability of holding a participant-owned account in SEED OK, (2) Hypothesis 2: SEED OK treatment increases participants' likelihood of holding a participant-owned account and (3) Hypothesis 3: there are interactive effects between an individual's financial knowledge and SEED OK interventions on account holding.
Data and Sample
SEED OK drew a probability sample of 7,328 children from all infants born in Oklahoma in two three-month periods in 2007: April through June and August through October. In selecting infants for the study, the SEED OK experiment oversampled racial and ethnic minority groups, including those who described themselves as African Americans, American Indian and Hispanic. Among 7,328 infants selected for the study from birth records, 213 cases were determined ineligible (e.g., because of the death of the infant or mother). Of the 7,115 remaining cases, the primary caregivers of 2,704 SEED OK children agreed to participate in the experiment and completed the baseline survey between fall, 2007 and spring, 2008, resulting in a participation rate of 38%. The 2,704 primary caregivers are considered the study participants. After the baseline survey, SEED OK randomly assigned 1,358 participants to the treatment group and 1,346 to the control group (Marks, Rhodes, and Scheffler 2008; Zager et al. 2010). A packet containing information about the OK 529 plan and the SEED OK experiment was sent to treatment participants by the OK Treasurer's Office after the random assignment.
A relatively low participation rate in SEED OK (38%) may raise the concern of non-response bias and could be related to the study's requirement of providing the infant's Social Security Number (SSN) to open an OK 529 account. However, SEED OK's participation rate is comparable to recent telephone surveys (e.g., 48% in the Survey of Consumer Attitudes in 2003 and 25% in Pew Research Center's 2003 national survey) (Curtin, Presser, and Singer 2005; Keeter et al. 2006). In addition, the comparison of study participants and non-participants using birth certificate data shows no significant differences for most observed characteristics between study participants and non-participants (e.g., infant's race and Hispanic origin, gender and birth weight; mother's marital status and metropolitan residency; and father's age), except for mother's age, education and nativity. However, differences among study participants and non-participants in these three characteristics are small: 25.53 years among study participants versus 25.22 years among non-participants for mother's age; 12.53 vs. 12.22 years for years of schooling; and 87% vs. 84% of native-born mothers. Furthermore, a weight variable was created to adjust the propensity of study participation using the observed information from birth records (Marks, Rhodes, and Scheffler 2008).
This study uses data from three sources in SEED OK: (1) birth records of SEED OK children, (2) a baseline survey conducted on all study participants through telephone interviews in the fall and spring of 2007–2008 and (3) quarterly account data obtained from the OK 529 plan. The timeline of data collection is provided in Table 1. Birth records include basic demographic and health information of SEED OK children and their biological parents. The baseline survey data contain detailed demographic and socioeconomic information of SEED OK participants, including financial knowledge. Finally, the quarterly account data provide accurate information (such as account-holding status, account balance, deposits and withdrawals) of participant-owned accounts opened for SEED OK children. SEED OK account data were released upon agreement with the OK 529 Board, and study participants were informed about this agreement (Nam et al. in press; Zager et al. 2010).
Table 1. Timeline of SEED OK Intervention and Data Collection
Intervention and Data Collection
Baseline survey conducted for SEED OK infants born in April through June 2007
State-owned accounts opened for infants born in April through June 2007
Baseline survey conducted for SEED OK infants born in August through October 2007
State-owned accounts opened for infants born in August through October 2007
Time-limited $100 account-opening incentive ended
Quarterly account data for the present study obtained
Follow-up survey conducted
Savings match ended
Of the 2,704 study participants, 2,651 are included in the final analysis sample. First, SEED OK participants who did not live in Oklahoma at the time of the baseline survey (n = 22) are excluded because non-residents in the control group may be less likely to open a participant-owned OK 529 account given the differences in tax benefits of the 529 plans in Oklahoma and their resident states. The final sample also excludes participants who are not mothers of SEED OK children (i.e., fathers, grandparents and siblings [n = 6]). This way, SEED OK children included in the analysis sample all have their mothers as study participants, and the characteristics of study participants are consistently measured (study participants are referred to as “mothers” hereafter). In addition, one case is removed because the SEED OK child died during the observation period. Finally, the study excludes 24 cases with missing values on several control variables used in the analytical models. These measures are discussed below.
The dependent variable—participant-owned account holding status—is created using information from the quarterly account data. This variable indicates whether a participant-owned account was opened and held by a SEED OK mother for a SEED OK child as of September 30, 2010. The OK 529 accounts opened and held by grandparents or other relatives of SEED OK children are not considered a participant-owned account in the study since SEED OK provides financial incentives and information only to mothers. The value of “1” is assigned to those with a participant-owned account on September 30, 2010 and “0” to others. This study does not examine other savings outcomes (e.g., amount of savings in participant-owned accounts) because only a small proportion of mothers have made deposits, and the distribution of savings is highly skewed.
The main independent variables are an indicator of SEED OK treatment status and a measure of financial knowledge. Mothers in the treatment group are coded as “1,” and those in the control group are coded as “0” for the treatment indicator. The indicator of treatment status, therefore, shows the institutional setting specific to each group to save for their children's college education. Financial knowledge is assessed using the information collected in the baseline survey with three questions drawn from the 2004 Health and Retirement Survey (Lusardi and Mitchell 2006). The three questions evaluate a mother's ability to understand basic financial concepts on compound interest, inflation and risk diversification:
“Suppose you had $100 in a savings account and the interest rate was 2% per year. After five years, how much do you think you would have in the account if you left the money to grow: more than $102, exactly $102, less than $102?”
“Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After one year, would you be able to buy more than, exactly the same as, or less than today with the money in this account?”
“Do you think that the following statement is true or false? Buying a single company stock usually provides a safer return than a stock mutual fund.”
The questions cover two content areas of financial knowledge proposed by Huston (2010)—money basics and investing. These questions have been adopted by multiple surveys as a measure of financial knowledge, including the 2004 Health and Retirement Survey, Wave 11 of the 1997 National Longitudinal Survey of Youth, the American Life Panel in 2008, and the 2009 Financial Capability Study (Lusardi and Mitchell 2011).
We created a dichotomous measure of financial knowledge based on mothers' responses to the three questions. Those who gave correct responses on all three questions are considered to have a high level of financial knowledge and are assigned the value of “1.” Others are assigned the value of “0.” Financial knowledge normally is indicated by an average of eight items that cover multiple content areas in existing research (Huston 2010). Although it has been commonly used in many studies about financial literacy (Lusardi and Mitchell 2011), the measure of financial knowledge used in this study is based on only three survey questions and, therefore, may be less reliable.
The study has four groups of control variables. First, children's characteristics include age (measured in months), gender (1 = male; 0 = female), and race. Taken from the birth record, child's race has five categories: non-Hispanic White, non-Hispanic African American, non-Hispanic American Indian, non-Hispanic Asian and Hispanic. The second group of control variables is mothers' characteristics (e.g., age, education [below high school, high school, some college and four-year college or above], marital status [1 = married, 0 = not married], and employment status [1 = employed, 0 = unemployed]).
Several measures of household characteristics (e.g., household size, number of children, homeownership [1 = homeowners; 0 = otherwise], welfare program participation [1 = yes; 0 = no] and household income-to-needs ratio) are included in the analyses. Household size is top-coded at “7” since only a small proportion of households have a household size greater than seven. The number of children is categorized into households with one child, two children, three or more children or a missing value, because nearly 40 mothers do not report this information in the baseline survey. Regarding welfare program participation, we assign the value of “1” to mothers whose households received income from TANF, Food Stamps, Supplemental Security Income (SSI) or Social Security Disability Insurance (SSDI) in the previous 12 months and the value of “0” to other participants. Household income-to-needs ratio is created by dividing self-reported pre-tax income in the previous 12 months by the 2007 US-Federal Poverty Guideline (http://aspe.hhs.gov/poverty/07poverty.shtml/). The Federal Poverty Guidelines are simplified poverty thresholds calculated for minimum levels of economic resources needed for different sizes of families. To address the missing values in the household income variable (n = 105), the sample is categorized into four groups according to the value of the household income-to-needs ratio: below two, between two and four, above four and missing values. Finally, the study includes a group of control variables indicating household asset-holding status (e.g., checking or savings accounts [0 = no; 1 = yes]; Certificates of Deposits [CDs], treasury bills, or corporate bonds [0 = no; 1 = yes; 2 = missing values]; savings bonds [0 = no; 1 = yes; 2 = missing values]; retirement accounts [0 = no; 1 = yes; 2 = missing values]; and other stocks or mutual funds [0 = no; 1 = yes; 2 = missing values]).
After descriptive analyses, we run two logit models since the dependent variable of participant-owned account holding is dichotomous. The first model regresses the dependent variable on two independent variables—financial knowledge and treatment status—and all the control variables. This model tests the first two questions of this study: the impacts of financial knowledge and the SEED OK intervention on one's chance of holding a 529 account for the child's future. In this model, the regression coefficient of financial knowledge indicates whether a higher level of financial knowledge increases the likelihood of account holding for SEED OK children regardless of treatment status. Similarly, the coefficient of treatment status shows the impact of the SEED OK experiment on 529 account holding.
In order to test the interactive effects of financial knowledge and the SEED OK intervention (the third question), the second model categorizes the sample into four groups: control group members with low-level financial knowledge, control group members with high-level financial knowledge, treatment group members with low-level financial knowledge, and treatment group members with high-level financial knowledge. The model can be expressed as follows:
where p(Yi) indicates the probability of holding a participant-owned account for mother i; (T0Fh)i denotes whether mother i belongs to the control group with high-level financial knowledge; (T1Fl)i denotes whether mother i belongs to the treatment group with low-level financial knowledge; (T1Fh )i denotes whether mother i belongs to the treatment group with high-level financial knowledge; and Xi is a vector of control variables.
β1, β2 and β3 in equation (1) are the parameters of interest and indicate differences in probabilities of holding a participant-owned account by mothers' level of financial knowledge within treatment group or control group. The coefficient of control group with high-level financial knowledge (β1) indicates how this group's chance of account holding is different from that of the control group with low-level financial knowledge (the reference group). Similarly, the difference between the other two coefficients (β2 and β3) assesses the difference in likelihood of account holding between those with low-level financial knowledge and those with high-level financial knowledge in the treatment group. This study uses a Wald test to investigate whether the difference between β2 and β3 is statistically significant. Using the results of equation (1), an approach proposed by Ai and Norton (2003) is used to evaluate the interactive effects between financial knowledge and treatment status:
where Ii is the interactive effect between financial knowledge and treatment status indicated by the change in the probability of participant-owned account holding for mother i; β1, β2, β3 and β4 are obtained from the estimation of equation (1), and Xi is a vector of control variables.
Supplementary analyses are conducted to test the robustness of our findings. First, two alternative measures of financial knowledge are created using the survey questions listed above. One measure sets the threshold of high-level financial knowledge at two rather than three correct responses. Following Lusardi and Mitchell (2006), this study also uses a continuous measure: the number of correct answers (0–3) from the three questions. Second, additional control variables of financial prudence and financial education experiences are added in analyses since these measures may be highly correlated with financial knowledge. Financial prudence is an aggregative measure from three financial management questions, including setting financial goals, sticking to a financial plan, and tracking spending. Financial education experiences measure whether participants have training on different financial strategies, such as budgeting, banking, credit and loans and investment. Third, we use complete-case analyses to remove observations with missing values in variables of household income, number of children and household asset-holding status (N = 2,423). All analyses in the study are weighted to take into consideration the oversampling of minority groups and potential non-response biases (Marks, Rhodes, and Scheffler 2008).
Table 2 reports characteristics of the study sample. The majority of SEED OK children are non-Hispanic White (65%). On average, SEED OK mothers are in their twenties, about 40% have at least some college education, and nearly two thirds are married. Given the young age of SEED OK children at the time of the baseline survey, it is not surprising that less than half of the mothers (46%) were employed at the time of the baseline survey. The mean household size is four, and two thirds of households have fewer than two children. Slightly less than half of the mothers own their homes, and four of every ten households received at least one public assistance benefit in the 12 months prior to the baseline survey assessment. Household income-to-needs ratio ranges from nearly zero to 44 times the federal poverty line. More than half of households (66%) have an income-to-needs ratio smaller than two. Less than one sixth of mothers (14%) has a high level of financial knowledge (i.e., answered all three financial knowledge questions correctly).
Table 2. Demographic and Socioeconomic Characteristics of the Sample (weighted)
Percentage or Mean
(N = 2,651)
(n = 1,341)
(n = 1,334)
Treatment status (treatment group)
Financial knowledge (high-level)
Count measure of financial knowledge (number of correct answers)
Age [mean by month (SD)]
Mother's (participant's) characteristics
Below high school
Four-year college or above
Marital status (married)
Employment status (employed)
Household size [mean (SD)]
Number of children
3 or more
Welfare participation (yes)
Household asset holding
Checking/savings accounts (yes)
CDs, treasury bills or corporate bonds
Stock or mutual funds
Treatment and control groups are comparable in terms of observed demographic and household characteristics listed in Table 2, columns 2 and 3. None of the characteristics is significantly different between the treatment and control groups. About 15% of mothers in both groups have a high level of financial knowledge, the average age of SEED OK mothers is about the same (26 years old), approximately 60% in both groups are married, the average household size is approximately four, and both groups have 66% of households with an income-to-needs ratio smaller than two.
Participant-owned Account Holding by Treatment Status and Financial Knowledge
Table 3 reports the percentage of mothers who hold a participant-owned account for SEED OK children by treatment status and financial knowledge. As of September 30, 2010, about 9% of mothers in the sample held a participant-owned account. While only 1% of control group participants had an account, the account-holding rate for the treatment group is 17%. The difference in the outcome measure between the two groups is statistically significant, suggesting that the CDA intervention has substantial impact on mothers' decisions to open and hold 529 accounts. Table 3 also indicates the impact of financial knowledge: 6% of mothers with a low level of financial knowledge hold participant-owned accounts, almost 17 percentage points fewer than those with a high level of financial knowledge. This comparison suggests that individual financial knowledge may be an important determinant of one's decision to hold a 529 account.
Table 3. Participant-owned Account Holding by Treatment Status and Financial Knowledge
POA Holding (%)
***p < .01
Participant-owned account holding in the full sample
Low-level financial knowledge
High-level financial knowledge
Financial knowledge by treatment status***
Control group: low-level financial knowledge
Control group: high-level financial knowledge
Treatment group: low-level financial knowledge
Treatment group: high-level financial knowledge
To examine the interaction of the SEED OK intervention and financial knowledge, mothers are further categorized into four groups by treatment status and level of financial knowledge. For the control group, those with a high level of financial knowledge have an account-holding rate (3.4%) about seven times that of those with a low level of financial knowledge (0.5%). The ratio, however, is reduced to less than four in the treatment group (45.0% and 11.9%).
Results of Logit Analyses
Results of Model 1
Results from Model 1 (see column 1 of Table 4) show that treatment status and financial knowledge have statistically significant coefficients after controlling for characteristics of child, mother and household. The results suggest that the SEED OK intervention significantly increases mothers' chances of holding participant-owned accounts for their children even when considering their level of financial knowledge (β = 3.51, p < .01). A calculation of the partial marginal effect of treatment status suggests that the SEED OK experiment increases the average predicted probability of holding a participant-owned account by 15.8 percentage points. At the same time, findings show the positive effect of financial knowledge on participant-owned account holding among mothers (β = .85, p < .01). If all SEED OK mothers had a high level of financial knowledge, the account holding rate in the sample would increase by 5.5 percentage points.
Table 4. Weighted Results of Logit Models on Participant-owned Account Holding (N = 2,651)
Note: OR, odds ratio. Only significant control variables are listed in the table.
*p < .1, **p < .05, *** p < .01.
Treatment status (ref: control group)
Financial knowledge (ref: low-level)
Treatment status and financial knowledge
Control group: low-level financial knowledge (ref.)
Control group: high-level financial knowledge
Treatment group: low-level financial knowledge
Treatment group: high-level financial knowledge
Child's gender (ref: female)
Child's race (ref: White)
Mother's education (ref: below high school)
Four-year college or above
Number of children (ref: 1)
Income-to-needs ratio (ref: below 200%)
Results of Model 2
The second model examines the interactive roles of SEED OK treatment status and financial knowledge. In this model, three dummy variables are included that differentiate the control group with low-level financial knowledge (the reference group) from the other three groups: the control group with high-level financial knowledge, the treatment group with low-level financial knowledge, and the treatment group with high-level financial knowledge.
Among control group members, financial knowledge does not make a significant difference in 529 account holding. Although the coefficient of the control group with high-level financial knowledge (β1 = 0.91, p = 0.25) is positive, it is not statistically significant at the 0.1 level, suggesting that participants' account holding does not differ by level of financial knowledge in the control group. Results also indicate that the SEED OK intervention has significant impacts: those in the treatment group are significantly more likely to hold an account than the reference group regardless of their level of financial knowledge (β2 = 3.55, p < .01; β3 = 4.39, p < .01). However, having financial knowledge makes a significant difference within the treatment group. The difference between the coefficients of treatment group with distinct levels of financial knowledge (β3 − β2 = 0.85, p < 0.01) is statistically significant at the 0.01 level, indicating that mothers with high-level financial knowledge are more likely to hold a participant-owned account than those with low-level financial knowledge in treatment group. The estimation of the partial marginal effect of financial knowledge in the treatment group shows that if SEED OK mothers in the treatment group all had a high level of financial knowledge, the account holding rate would increase by 10 percentage points.
We apply equation (2) to calculate the interactive effects between financial knowledge and treatment status. The result shows that the probability difference in account holding between those with high-level and low-level financial knowledge in the treatment group is 8.6 percentage points (p < .1) higher than the probability difference between those with high-level and low-level financial knowledge in the control group. The hypothesis of interactive effects is supported.
Figure 2 displays the interactive effects by reporting the predicted probabilities of holding a participant-owned account for a typical SEED OK child in different scenarios. Using the median values of control variables, a typical case is defined as a 41-month-old White male child whose mother is 25 years old, married and unemployed and has a high school degree; whose household has four members (including two children) with income lower than 200% of the poverty line; whose household does not receive any public assistance; and who does not own a home or any other types of financial assets other than a checking/savings account. As reflected by the predicted probabilities for the control group in the existing policy context, the probability of having an OK 529 plan account for a typical participant is extremely small regardless of participant level of financial knowledge (0.25% and 0.59%). The SEED OK treatment increases the predicted probabilities of account holding to 7.9% and 16.6%, respectively, for those with low and high levels of financial knowledge. The findings on the interactive effects imply that financial knowledge may help individuals take advantage of financial incentives and utilize financial information provided by the SEED OK experiment.
Results on Control Variables
Models 1 and 2 have the same results on control variables. Five control variables are statistically significantly associated with the outcome measure. First, male children are statistically more likely to have a participant-owned account held for them than female children (odds ratio = 1.6), although it is not clear why. Second, compared to non-Hispanic White infants, non-Hispanic African American infants and non-Hispanic American Indian infants are less likely to be beneficiaries of participant-owned accounts. Third, there is a positive association between mothers' college education and the probability of holding a participant-owned account. In addition, households with three or more children and those with an income-to-needs ratio greater than four are more likely to hold an account for their children. Due to the missing values in control variables (e.g., income-to-needs ratio, number of children and asset-holding status), several missing indicators have been included in the analyses. All of these missing indicators are not statistically significant, suggesting that there is not different propensity of holding a 529 account between those with a missing value and the reference group (e.g., missing in income-to-needs ratio vs. income-to-needs ratio below two, and missing in number of children vs. one child).
Robustness tests (described in the Methods section) have substantively identical results as those reported above for almost every analysis. One exception is that when a high level of financial knowledge is defined by two instead of three correct responses among the three survey questions, financial knowledge loses its statistical significance at the 0.1 level in Model 1.
Hypothesis of Financial Knowledge
The hypothesis of financial knowledge is supported by analysis results in this study. As shown in results from the first logit model, mothers with a high level of financial knowledge are more likely to hold a participant-owned account when controlling for treatment status and other socioeconomic characteristics. The average marginal effect of financial knowledge in the sample is a 5.6-percentage-point increase in account-holding rate. Also, in the treatment group, the predicted probability of holding a 529 account is 16.6% for a typical participant with a high level of financial knowledge and 7.9% for a participant with a low level of financial knowledge (see Figure 2).
The finding regarding financial knowledge is consistent with the literature on financial literacy. Existing studies of the effects of financial literacy show that financial knowledge is positively correlated with financial well-being, such as retirement planning and savings, pension contributions and overall wealth (Behrman et al. 2010; Lusardi and Mitchell 2011). Individuals with high-level financial knowledge are expected to display “good financial behavior” (Hung, Parker, and Yoong 2009) and appropriate financial decision making and planning (Remund 2010). This study adds to the literature by focusing on financial behavior related to college savings. Since long-term financial planning for college is considered desirable financial behavior, a positive association between financial knowledge and 529 account holding is expected.
The positive association between financial knowledge and 529 account holding is consistent with the finding from a study of the Maine 529 plan (Huang, Beverly, Clancy, Lassar, and Sherraden in press), which shows that financially sophisticated parents are more likely to open 529 accounts for their children. The Maine study, however, does not have a direct measure of financial knowledge and uses stock/bond ownership as a proxy for financial sophistication. This study shows that the measure of stock ownership is not statistically significant when a direct measure of financial knowledge is present.
Hypothesis of Policy Experiment
Results also suggest that the SEED OK experiment increases participant-owned account holding among mothers. As of September 30, 2010, the participant-owned account-holding rate was 17% in the treatment group but only 1% in the control group. Being assigned to the treatment group increases the odds of holding an account 33.5 times. While enrollment even in the treatment group is far from universal, the difference in account-holding rates between the two groups is substantial. In comparison, the Maine 529 program enrollment rate was 10% for one-year-old children in 2009 when a $500 financial incentive was offered to every newborn in the state (Huang et al. in press). Findings in the current study support a proposition of the financial capability concept: the institutional setting (e.g., incentives, information and access) is a critical factor in achieving financial well-being.
Several components of the SEED OK experiment, especially those targeting participant-owned accounts listed in Figure 1, may contribute to the higher account-holding rate in the treatment group. The $100 time-limited account-opening incentive for the treatment group is likely to motivate mothers to open a participant-owned 529 account. For income-eligible mothers, holding a participant-owned account also is necessary to receive matching funds for individual savings. The financial incentive ($1,000) in state-owned accounts and the program information package sent by the Treasurer's Office also may increase awareness of the OK 529 plan among mothers, facilitating opening and holding of a 529 account. Other policy innovations (e.g., the Alfond program in Maine) also report a positive association between financial incentives and 529 account holding (Huang et al. in press).
Hypothesis of Interactive Effects
Model 2 shows that mothers' level of financial knowledge does not affect participant-owned account holding among the control group. High-level financial knowledge significantly increases the probability of account holding only in the treatment group. It seems that high-level financial knowledge motivates mothers in the SEED OK treatment group to respond to the policy incentives more actively by opening and holding a participant-owned account. Financial knowledge matters only in the treatment group probably because mothers have barriers that are too challenging to overcome without institutional support. Most financially knowledgeable mothers are not motivated enough to open a 529 account without strong incentives probably because they are too busy taking care of an infant or because they believe they do not have enough money to save for a long-term goal. To place this finding in context, there is a very low 529 plan participation rate in the general population: Less than 4% of all Oklahoma children under age 18 are named as the beneficiary of an OK 529 account.
These findings support the third proposition that the association between financial knowledge and college saving is partially defined by institutional features, an observation which is expected in SEED OK. The effectiveness of the SEED OK experiment may rely on participants' understanding and perception of financial incentives and other features, which are largely related to financial knowledge. From the perspective of financial capability, financial knowledge and features of the SEED OK experiment work interactively to shape a participant's financial decisions. In other words, financial knowledge helps participants explore and understand institutional features, while simultaneously, various institutional features create opportunities for participants to apply their financial knowledge.
Direction of Interactive Effects
Results of Model 2 seem to reflect a general direction of interactive effects: the more incentives the institutional features provide, the greater the impact of financial knowledge on financial well-being. However, this is not always true because the direction of interactive effects can be shaped by specific policy features. For instance, the direction of interactive effects might change if a program provides information or simple choices that an individual with low-level financial knowledge can understand or if a desirable financial decision does not rely on an individual's financial knowledge.
Although not an outcome measure of this study, state-owned accounts in SEED OK are a perfect way to examine the direction of interactive effects. All but one participant in the treatment group hold automatically opened state-owned accounts for their child. Compared to the holding rate of participant-owned accounts, the state-owned account option achieves the goal of universal accounts for children. If we apply Model 1 to the outcome measure of state-owned accounts rather than participant-owned accounts (we cannot run a multivariate regression because of the lack of variance in this outcome measure), treatment status would explain all variance on account holding. Financial knowledge does not have a significant effect on the outcome measure even in the treatment group, and the impact of individual financial knowledge on holding a state-owned account becomes negligible.
In summary, the direction of interactive effects is defined by the institutional features in SEED OK. When individual actions (i.e., opening and holding a participant-owned account) are required to take advantage of financial incentives in SEED OK, a higher level of financial knowledge leads to a higher account-holding rate. When automatic opening is provided for state-owned accounts, individuals' financial knowledge does not matter.
The findings of this study have several policy implications. First, people with high-level financial knowledge are more likely to make sound financial decisions and display desirable financial behaviors than those with low-level financial knowledge. In SEED OK, financial knowledge is positively related to participant-owned account holding in the treatment group. As is widely suggested by the literature on financial literacy, appropriate financial education and training should be provided for individuals to improve financial knowledge. Existing research shows some evidence of effectiveness of financial education (Grimes, Rogers, and Smith 2010; Walstad, Rebeck, and MacDonald 2010). In the future, it will be important to study the effectiveness of financial education curricula and teaching methods. Such research should be conducted with different populations, especially in populations of low-socioeconomic status.
However, results also suggest that financial education and training alone is not sufficient to expand financial capability. Equally as important as financial education is the creation of a supportive institutional setting. In SEED OK, policy affects the association between financial knowledge and participant-owned account-holding status, and the influence of financial knowledge is magnified in the treatment group.
A supportive institutional setting—including access to financial education and financial services—can expand financial capability. Several studies (e.g., Sherraden in press; Sherraden, Schreiner, and Beverly 2003) identify key constructs (e.g., incentives, information and access) and guidelines that can be used to create supportive institutional settings and shape individual financial actions. Moreover, findings from the field of behavioral economics provide guidance about promising design features (e.g., Thaler and Sunstein 2008).
Because there is no single financial services strategy applicable in all situations and for all target populations, different policy innovations should be encouraged and tested to achieve a supportive policy environment. In SEED OK, for example, it is highly unlikely that additional financial incentives will by themselves achieve universal enrollment. However, automatically opened state-owned accounts, a different policy feature, result in a nearly 100% participation rate in the treatment group.
This study examines the effect of financial knowledge on 529 account holding in the SEED OK experiment, and the findings support the propositions of financial capability. Financial knowledge increases the participant-owned account-holding rate in the treatment group but not in the control group. The interactive effects between financial knowledge and treatment status are statistically significant. This indicates that the association between financial knowledge and financial decision making varies depending on the policy context. In SEED OK, the policy experiment seems to have greater impacts on the participant-owned account-holding rate than financial knowledge of an individual. From a public policy perspective, both effective financial education and supportive institutional features should be considered to expand financial capability.